To the Editor.
—The recent article by Roy et al1 describes an impressive retrospective study quantifying the error rate of human immunodeficiency virus (HIV) serum antibody testing in a large cohort of military personnel. The main conclusion of the article is that "true seroreversion" (ie, loss of confirmed HIV seropositivity not resulting from clearance of passively acquired maternal antibodies) does not occur, and that "in fact, most (if not all) cases of apparent seroreversion represent errors of attribution or testing." We feel that it is important not to overinterpret these results or necessarily to extend the conclusion to other nonmilitary populations.More than 2.5 million individuals were tested in this study, and 4911 were confirmed HIV-seropositive. Only six individuals (excluding 31 transiently seropositive infants) received at least two positive tests followed by a negative test and met the authors' definition of "potential seroreverters." It would have been useful to